IEICE Technical Committee Submission System
Conference Paper's Information
Online Proceedings
[Sign in]
Tech. Rep. Archives
 Go Top Page Go Previous   [Japanese] / [English] 

Paper Abstract and Keywords
Presentation 2022-01-18 14:00
Robustness to Adversarial Examples by Mixtures of L1 Regularazation Models
Hironobu Takenouchi, Junichi Takeuchi (Kyushu Univ.) IBISML2021-26
Abstract (in Japanese) (See Japanese page) 
(in English) We propose a method of adversarial training using L1 regularizationfor image classification.It is known that L1 regularization is effective to improve robustness for adversarial samples by pruning image classifier's explanatory variables.However, we showed by experiments that the regularized model is robust to the adversarial samples to the model without L1 regularization,while it has vulnerability to the adversarial samples generated using the knowledgeabout the L1 regularized model.For this problem, we developed a new method using amixture of models with various strength of L1 regularizationand trained it with adversarial samples to single L1 regularized models.By experiment, we showed that our model is more robust than the single modelsand that the successful adversarial perturbation to the mixture modelis less diverse than that to other single models.
Keyword (in Japanese) (See Japanese page) 
(in English) Deep Learning / Adversarial Training / / / / / /  
Reference Info. IEICE Tech. Rep., vol. 121, no. 321, IBISML2021-26, pp. 61-66, Jan. 2022.
Paper # IBISML2021-26 
Date of Issue 2022-01-10 (IBISML) 
ISSN Online edition: ISSN 2432-6380
Copyright
and
reproduction
All rights are reserved and no part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Notwithstanding, instructors are permitted to photocopy isolated articles for noncommercial classroom use without fee. (License No.: 10GA0019/12GB0052/13GB0056/17GB0034/18GB0034)
Download PDF IBISML2021-26

Conference Information
Committee IBISML  
Conference Date 2022-01-17 - 2022-01-18 
Place (in Japanese) (See Japanese page) 
Place (in English) Online 
Topics (in Japanese) (See Japanese page) 
Topics (in English) Machine Learning, etc. 
Paper Information
Registration To IBISML 
Conference Code 2022-01-IBISML 
Language Japanese 
Title (in Japanese) (See Japanese page) 
Sub Title (in Japanese) (See Japanese page) 
Title (in English) Robustness to Adversarial Examples by Mixtures of L1 Regularazation Models 
Sub Title (in English)  
Keyword(1) Deep Learning  
Keyword(2) Adversarial Training  
Keyword(3)  
Keyword(4)  
Keyword(5)  
Keyword(6)  
Keyword(7)  
Keyword(8)  
1st Author's Name Hironobu Takenouchi  
1st Author's Affiliation Kyushu University (Kyushu Univ.)
2nd Author's Name Junichi Takeuchi  
2nd Author's Affiliation Kyushu University (Kyushu Univ.)
3rd Author's Name  
3rd Author's Affiliation ()
4th Author's Name  
4th Author's Affiliation ()
5th Author's Name  
5th Author's Affiliation ()
6th Author's Name  
6th Author's Affiliation ()
7th Author's Name  
7th Author's Affiliation ()
8th Author's Name  
8th Author's Affiliation ()
9th Author's Name  
9th Author's Affiliation ()
10th Author's Name  
10th Author's Affiliation ()
11th Author's Name  
11th Author's Affiliation ()
12th Author's Name  
12th Author's Affiliation ()
13th Author's Name  
13th Author's Affiliation ()
14th Author's Name  
14th Author's Affiliation ()
15th Author's Name  
15th Author's Affiliation ()
16th Author's Name  
16th Author's Affiliation ()
17th Author's Name  
17th Author's Affiliation ()
18th Author's Name  
18th Author's Affiliation ()
19th Author's Name  
19th Author's Affiliation ()
20th Author's Name  
20th Author's Affiliation ()
Speaker Author-1 
Date Time 2022-01-18 14:00:00 
Presentation Time 20 minutes 
Registration for IBISML 
Paper # IBISML2021-26 
Volume (vol) vol.121 
Number (no) no.321 
Page pp.61-66 
#Pages
Date of Issue 2022-01-10 (IBISML) 


[Return to Top Page]

[Return to IEICE Web Page]


The Institute of Electronics, Information and Communication Engineers (IEICE), Japan